jgagneastro/coffeegrindsize

Detects the individual coffee grounds in a white-background picture to determine particle size distribution

coffeegrindsize's Issues

Version for Ms Window

Any version of the software can be run in Microsoft Windows?

Mac installation instructions can be simplified

The installation guide describes the "hard way" to get around the "unidentified developer" issue. It's easier to just Ctrl-click (or right-click) the app in Finder and select Open from the shortcut menu.

https://support.apple.com/guide/mac-help/open-a-mac-app-from-an-unidentified-developer-mh40616/mac

Obviously, this doesn't need to be fixed, but I just wanted to point it out. I am also a developer of an "evil" non-registered Mac app :-)

Made easy instructions for Korean. (한국어 간편 사용설명서 제작)

I found this project very useful, but hard to use for non-english speakers. So I made rough user guide for other Korean fellows.

간단하게 한국어 사용설명서를 만들었습니다.

https://github.com/im-hjk/coffeegrindsize

Source distribution

It looks like we have a packaged distribution of this tool here. Can you please include the source files that produced the distribution? You mentioned that pull requests were welcome, and providing the source in that form would make it easier to do that.

Starting a Community Database of Grinder particle size distributions

While I think this is an awesome tool, the real value would be in gathering and aggregating as much of this sort of data together as a community. If we are able to share all our grind distributions from all of our own grinders and compare them to each other, then we would be able to use the data to answer some often asked questions such as:

Is grinder X comparable or better than grinder Y?
Is it true that beans from origin A produce more fines than those from origin B?
What effect does the roast level, days past roast, etc. have on the grind quality?

I created a sample repo with my own data, but it would be better if you can approve/administrate this project with your wider reach and influence.

https://github.com/Coffee-Grind-Distribution/coffee-grind-distribution

Particle Detection Fails

Running on Arch Linux, Particle Detection fails with the following error:

Exception in Tkinter callback
Traceback (most recent call last):
  File "/usr/lib/python3.9/tkinter/__init__.py", line 1892, in __call__
    return self.func(*args)
  File "/home/user/coffeegrindsize/coffeegrindsize.py", line 592, in <lambda>
    psd_button = Button(toolbar, text="Launch Particle Detection", command=lambda: self.launch_psd(None),highlightbackground=toolbar_bg)
  File "/home/user/coffeegrindsize/coffeegrindsize.py", line 2341, in launch_psd
    self.cluster_data.append(clusteri_data)
AttributeError: 'NoneType' object has no attribute 'append'

Histogram cannot be created

I'm not able to create a histogram after Particle Detection step:

Exception in Tkinter callback
Traceback (most recent call last):
  File "/usr/lib/python3.9/tkinter/__init__.py", line 1892, in __call__
    return self.func(*args)
  File "/home/ghotrix/.local/lib/python3.9/site-packages/coffeegrindsize/scripts/coffeegrindsize.py", line 600, in <lambda>
    histogram_button = Button(toolbar, text="Create Histogram", command=lambda: self.create_histogram(None), highlightbackground=toolbar_bg)
  File "/home/ghotrix/.local/lib/python3.9/site-packages/coffeegrindsize/scripts/coffeegrindsize.py", line 2973, in create_histogram
    figdata = figdata.reshape(fig.canvas.get_width_height()[::-1] + (3,))
ValueError: cannot reshape array of size 5298750 into shape (1130,1000,3)

.

Very small particles are not detected

Do you have a tip how to change the thresholds in a way that also the very small particles (blue circles) are counted?

Sourcefile (jpg since png is too large to upload)

coffeegrindsize quit unexpectedly

Can't open the app on Mac with BigSur 11.1 (security settings have been changed, not the issue)

FEATURE REQUEST: Use instance segmentation for particle detection

The particle detection could be automated (and user to user bias removed) with an instance segmentation algorithm instead of thresholding. Given enough time and images, i can implement and train that if it's desired. I can fork off and start if 1) you think it's a good idea, and 2) there are a lot of images (hopefully with their threshold settings noted) somewhere or someway to get them. What do you think?